Should Access make a P4-1.6 run 100% during a query when a P120 could do the same job before?

[MadRat]

At work we switched out a P120 workstation with a P4-1.6 (nope, not a Northwood 🙁) processor. Funny how one of the hardest stresses on this machine is a simple Access query pulled from a database across the network. Hardly a real stress to a CPU normally, but on this system its really struggling. The chipset is an i845 with (gasp!) SDRAM. (I know, I know, but purchasing this slug wasn't my choice!) Seems like a CPU shouldn't necessarily need to run 100% when its a simple query across a 10bT connection. Is the P4 this poor at sorting that it cannot handle the task when a P120 could??

 

[ScrewFace]

Nope. It's just the Access program. It always asks 100% from your CPU no matter what it is.

 

[MadRat]

I kind of figured as much. Pretty screwy of Microsoft to do that. 😛

 

[MikeMike]

if its important, to be able to use it for other stuff, and your using xp or something similar, set its prioriy to low or realtime, which will allow other programs to take over. IF low is to slow, set it to realtime, which will allow other programs to take over cpu% and once they are done it will go back up.

I think i just made an ass of myself

 

[Gstanfor]

This is one of those jobs where you get to see just how badly Intel conned you with their netburst architecture, actually.

For a Pentium4 to dowell in a task it has to keep as much relevant data as it possibly can in its caches and avoid branching as much as possible or its pipelines suffer badly.

Access queries are all about random data accesses and branching around, so they are a nightmare task for the P4. You would do much better running access on a P3 system or an Athlon. Access turns netburst into netbusted.

Greg

 

[MadRat]

The P4's seem to benchmark extremely well in non-optimized sorting and archiving tasks. Why would Access make it suffer?

 

[Gstanfor]

The P4's seem to benchmark extremely well in non-optimized sorting and archiving tasks. Why would Access make it suffer?

I don't know what specific apps you are talking about there. If you are referring to benchmarks, they are rigged to make the P4 look good. Data archiving shouldn't be as much of a problem (if you are talking about creating zips etc) since all you are really doing is packing data from expected locations into a contiguous file.

What kills access performance is the random nature of queries. the next memory location to be searched may be totally different to the previous one. This makes caching the data almost impossible. also there is the branching caused by the query rules going on.

the effect this has on the P4 is that it cannot keep its pipeline full or is being forced to constantly flush the pipeline, and as a result, performance plummets (keeping the P4's pipeline full is critical to getting good performance out of it and anything that intensively processes memory randomly or semi-randomly, or branches around a lot is going to make the P4 suffer badly).

Greg

 

[BFG10K]

Pretty screwy of Microsoft to do that. 😛

Uh, no.

Usually programs that are required to work as fast as possible (eg games, number crunching, etc) will take all the CPU cycles they can get. Those that only require a certain threshold of CPU cycles (eg playing MP3s) will not take more when they reach the target.

 

[BFG10K]

And obviously the query will finish faster on the P4 system as well.